gcc.dg/autopar test regressions

[Andrew Stubbs]

Linaro GCC has some test regressions vs. FSF:

armel only:

+FAIL: gcc.dg/autopar/reduc-1.c scan-tree-dump-times parloops "Detected reduction" 3
+FAIL: gcc.dg/autopar/reduc-1.c scan-tree-dump-times parloops "SUCCESS: may be parallelized" 3
+FAIL: gcc.dg/autopar/reduc-1char.c scan-tree-dump-times parloops "Detected reduction" 3
+FAIL: gcc.dg/autopar/reduc-1char.c scan-tree-dump-times parloops "SUCCESS: may be parallelized" 3
+FAIL: gcc.dg/autopar/reduc-1short.c scan-tree-dump-times parloops "Detected reduction" 3
+FAIL: gcc.dg/autopar/reduc-1short.c scan-tree-dump-times parloops "SUCCESS: may be parallelized" 3
+FAIL: gcc.dg/autopar/reduc-2.c scan-tree-dump-times parloops "Detected reduction" 3
+FAIL: gcc.dg/autopar/reduc-2.c scan-tree-dump-times parloops "SUCCESS: may be parallelized" 3
+FAIL: gcc.dg/autopar/reduc-2char.c scan-tree-dump-times parloops "Detected reduction" 2
+FAIL: gcc.dg/autopar/reduc-2char.c scan-tree-dump-times parloops "SUCCESS: may be parallelized" 2
+FAIL: gcc.dg/autopar/reduc-2short.c scan-tree-dump-times parloops "Detected reduction" 2
+FAIL: gcc.dg/autopar/reduc-2short.c scan-tree-dump-times parloops "SUCCESS: may be parallelized" 2
+FAIL: gcc.dg/autopar/reduc-6.c scan-tree-dump-times parloops "FAILED: it is not a part of reduction" 3

[CodeSourcery internal ID: #6663]

[Yao Qi (yao-codesourcery)]

This problem is caused by 'cunroll pass unrolls the loops so that parloops pass has no loop to parallelize'.

Tree dump of cunroll show the difference,
on FSF GCC 4.4 trunk (built as arm-unknown-linux-gnueabi),
Loop 1 iterates 15 times.
  Loop size: 7
  Estimated size after unrolling: 32
Not unrolling loop 1. <<----

on linaro GCC (build as arm-unknown-linux-gnueabi)
Loop 1 iterates 15 times.
  Loop size: 7
  Estimated size after unrolling: 32
Removing basic block 48 <<---

The flag of cunroll is set in tree-ssa-loop.c:tree_complete_unroll(),

 return tree_unroll_loops_completely (flag_unroll_loops
           || flag_peel_loops
           || optimize >= 3, true);

on FSF GCC 4.4, flag_unroll_loops is 0, however in linaro gcc, flag_unroll_loops != 0. flag_unroll_loops is set to a non-zero value in gcc/config/arm/arm.c: arm_optimization_options() and arm_override_options().

That is reason why there is no such problem when linaro gcc is built natively on x86. Looks like root cause is found, but have to figure out how to fix it in a reasonable way.

[Viktor (vchong)]

GCC version no longer supported.